summaryrefslogtreecommitdiff
path: root/tokenizer/bpe
diff options
context:
space:
mode:
authorJonas Knobloch <jonas.knobloch@t-online.de>2026-09-09 02:41:24 +0200
committerJonas Knobloch <jonas.knobloch@t-online.de>2026-09-09 02:41:24 +0200
commit3c5f4c4279e6c9d6479e40eff4fbe1add59c1b3a (patch)
treec12ce174c117b0205bc3426b6cc2501f7e709f92 /tokenizer/bpe
parent8fd3f8aeec85f47d43ef36a81456d48640ca33f9 (diff)
WIP
Diffstat (limited to 'tokenizer/bpe')
-rw-r--r--tokenizer/bpe/utility.go4
1 files changed, 4 insertions, 0 deletions
diff --git a/tokenizer/bpe/utility.go b/tokenizer/bpe/utility.go
index f8fcdbc..79e6faf 100644
--- a/tokenizer/bpe/utility.go
+++ b/tokenizer/bpe/utility.go
@@ -19,6 +19,10 @@ func NewTokenizerFromFiles(vocab, merges string, cfg Config) (*Tokenizer, error)
return NewTokenizer(tokenizer, cfg), nil
}
+func MBPE(t *Tokenizer) *mbpe.Tokenizer {
+ return t.mbpe
+}
+
func Vocab(t *Tokenizer) []string {
m, ok := t.mbpe.Model().(*mbpe.MBPE)